【作者】 何俊超；

【导师】 齐乐华；

【作者基本信息】 西北工业大学 ， 机械设计及理论， 2003， 硕士

【摘要】 复合材料管、棒材液-固挤压成形工艺是近几年发展起来的一种金属成形新工艺，具有工序少、成本低、制件性能好等特点，应用前景广阔，但该工艺是一多变量非线性时变系统，很难建立精确的数学模型，过程参数难于控制，限制了该工艺的实际应用。 本文通过大量的试验，对液-固挤压成形工艺中的影响因素进行了研究，初步掌握了该成形工艺的一般规律；在分析实验数据的基础上确定了实验样本集，并初步确定了该工艺系统的模糊推理规则；利用模糊神经网络技术对液-固挤压工艺系统进行建模，实现了对样本数据的良好拟合；补偿模糊神经网络在训练过程中能够根据样本动态调整网络参数，提取较为准确的模糊推理规则；通过利用该工艺系统模型进行参数预测，得到了与实验数据吻合较好的预测结果。 遗传算法在优化过程中不需要求导或其它辅助知识，只需要影响搜索方向的目标函数和相应的适应度函数，对难于建立数学模型的非线形系统有较强的优化能力。本文根据液-固挤压成形工艺的特点，利用遗传算法建立了工艺参数优化模型、进行了工艺参数优化，结果比较理想。 通过对复合材料液-固挤压成形工艺系统建模的研究，探讨了模糊神经网络参数的物理意义及其对网络学习速度的影响，确定了可加快网络学习速度的参数赋值方法；结合液-固挤压工艺特点探讨了利用遗传算法进行参数优化过程中一些问题的处理方法及遗传操作算子对优化过程的影响，确定了参数选取的合理范围。 本文利用Microsoft VC 6.0、Matlab以及Microsoft Access等软件，自行编制了一套基于模糊神经网络与遗传算法的集工艺建模、参数预测、参数优化、图形输出及数据库访问为一体的液-固挤压成形工艺应用软件，实现了该软件的可视化及良好的交互性。通过该软件可实现工艺参数的预测及优化，为液-固挤压工艺过程参数的正确选取及实际应用奠定了基础。更多还原

【Abstract】 Liquid-solid extrusion of composite material is a new kind of metal forming process, which has been developed in recent years with a promising practical application prospect for its simple working procedure, low cost and good workpiece performance. But it is difficult to set up an accurate mathematical model because the process is a multivariable nonlinear time varying system. So this process can’t be put into practical uses for the difficulties in process parameters control.In this paper, some key factors in the process of composite liquid-solid extrusion had been researched through a large amount of experiments and the common rules of composite liquid-solid extrusion had been obtained. The sample data was built and the common fuzzy rule was set up through analyzing the experimental data collected during the experiments. The fuzzy neural network model of composite liquid-solid extrusion had been set up, the sample data had been fitted and the accurate fuzzy rule had been concluded through training the fuzzy neural model with the sample data. The process parameters had been predicted by use of the fuzzy neural network model constructed, and the predicted results coincide well with the experimental data.Optimization can be conducted by use of genetic algorithms. This optimization method needn’t the differential coefficient and other assistant knowledge, but of the object function and the adaptive evaluation function. It’s powerful to optimize the nonlinear system difficult to set up the mathematical model. In this paper, the composite liquid-solid extrusion optimization model is built by means of genetic algorithm. The process parameters had been optimized and the ideal optimization results are obtained.The physical meaning of the fuzzy neural network parameters and their influence on the model performance are discussed through the research of composite liquid-solid extrusion model, and the reasonable method of evaluating the network parameters is confirmed which can quicken the training speed. Some problems and the influence of the evolution operator on the optimization course are discussed, arid the reasonable range of the evolution parameter is concluded.In this paper, a set of composites liquid-solid extrusion application software is complied by using of the Visual C++, Matlab and Access software. The visualization and interaction are realized through this software, what’s more, the functions of fuzzy neural model setting up, parameter prediction, parameter optimization, figureexporting and database accessing are included. The composite liquid-solid extrusion process parameters can be predicted and optimized by use of the software. The author’s work lay a foundation for the reasonable choose of the process parameters and practical application of the composite liquid-solid extrusion.更多还原